The bulk of our linguistic activities, be they worldly or academic, is interactive: we use language, in a variety of modalities, to connect and communicate with interlocutors who may be nearby or spatio-temporally removed from us. In turn, much of our linguistic interaction with others – be it in the capacity of speakers, hearers or onlooking audiences – involves, and originates in, disagreement – broadly understood as a conflict between two or more parties’ mental states (e.g. propositional beliefs). This extremely broad characterisation captures a vast class of situations; but a very natural distinction may be drawn between disagreements that are about the world, on the one hand, and about language, on the other. The world is at stake when Giulia maintains that the Ottoman Empire carried out a genocide of the Armenian people in the early 20th century, and a Turkish state authority denies this. And language is at stake when Pedro and Erich disagree over the meaning of the term ‘metalinguistic disagreement’. At a first approximation, then, we may say that linguistic agents disagree not only through language, but also about language.

The central topic of this collection falls on the latter side of the divide: a metalinguistic disagreement, at its most general, is any disagreement about the meaning or adequate usage of some linguistic expression. This is still incredibly broad, of course, and somewhat opaque. A first helpful step, then, is to distinguish descriptive from normative metalinguistic disagreements. Descriptive metalinguistic disagreements involve conflicting views on the actual meaning of some term. Take, for instance, the conflict over the meaning of ‘metaphysics’ between a philosophy teacher and a student interested in the role of quartzes in spiritual healing and adventuring; or the disagreement over the meaning of ‘concept’ between, say, a cognitive scientist and a marketing analyst. Such possibilities of linguistic misalignment, and consequent failures of communication, have long been the subject of philosophical interest (e.g. Locke 1979, James 1907, Carnap 1963, Sidelle 2007, Hirsch 2005, 2009, Chalmers 2011).

In turn, normative metalinguistic disagreements come about when two or more parties entertain conflicting views as to the meaning that some term ought to have. Controversies about whether the definition of ‘sport’ should be broad enough to include chess and videogames, about whether catcalling should be counted in the extension of ‘sexual violence’, about what gets to be called ‘bio’, or over the definition of ‘unbearable suffering’ in assisted death regulation fall into this category. What these examples, and the theoretical notion they instantiate, bring to light is that the words we choose, the meanings we associate with those words, and the concepts we have ready to hand matter – sometimes, to life-changing extents. Such choices inescapably shape our lives and projects, theoretically and also practically speaking, at an individual or a social level. Increased attention towards these aspects of our linguistic activities is reflected in the recent explosion of discussions on conceptual ethics and conceptual engineering (see e.g. Haslanger 2012; Burgess and Plunkett 2013a, b, Plunkett & Sundell 2013a, Ludlow 2014, Cappelen 2018, Cappelen et al. 2020, Marques and Wikforss 2020).

Two additional distinctions complete the basic anatomy of the notion. First, metalinguistic disagreements may – but need not – be expressed in a dispute (Cappelen and Hawthorne 2009; MacFarlane 2014). Second, when metalinguistic disagreements come to be expressed in some dispute, this may happen explicitly or implicitly. Explicit expressions of metalinguistic disagreements mention the relevant expression, and are transparent about the representational-level nature of the disagreement. Pedro’s utterance “We should save the term ‘metalinguistic’ to refer to disputes,’’ and Erich’s counter-utterance “No, the term ‘metalinguistic’ should be allowed to encompass unexpressed disagreements about language, disagreements taking place at the meta-level’’ jointly count as an explicit expression of Pedro and Erich’s disagreement over the meaning of the term ‘metalinguistic’.

In some cases, by contrast, metalinguistic disagreements reach the verbal surface only implicitly, via conflicting uses of some disputed term. So-called merely verbal disputes can be thought of as externalisations of implicit descriptive disagreements. When Ana asserts “Bruno lives by the bank” and Claudia responds “No, he doesn’t live by the bank”, their sole disagreement concerns the meaning associated with ‘bank’, not Bruno’s address (Hirsch 2005, 2009; Sidelle 2007; Chalmers 2011; Jenkins 2014; Balcerak Jackson 2013, 2014; Belleri 2017, 2018; Vermeulen 2018). Implicit normative disagreements, on the other hand, give rise to metalinguistic negotiations (Plunkett and Sundell 2013a). When David utters “Waterboarding is torture” and Edgar responds “No, waterboarding is not torture”, their disagreement is not (immediately) about torture but rather about (the meaning of) the term, ‘torture’. Moreover, the disagreement between David and Edgar may plausibly be construed as a conflict over what the term ought to mean, rather than about what it actually means.

Notice that the idea that normative divergences about language are sometimes disguised as standard object-level disagreements is not entirely new. A famous precursor is found in Rudolf Carnap’s (1950) suggestion that disputes over the existence of abstract entities be reinterpreted as disagreements about whether to adopt some particular linguistic framework. More recently, Schiappa (2003) and Ludlow (2014) have applied similar analyses to a number of both theoretical and everyday discussions. An especially prominent articulation of the metalinguistic analysis of disagreements has been given by Plunkett and Sundell (2013a), resulting in a rich and lively debate over, both, the theoretical details of the proposal (Plunkett 2015; Kocurek et al. 2020, Mankowitz 2021, Plunkett & Sundell 2021a, b, Rast 2022, Soria Ruiz 2023), and its applications in multiple areas of philosophical interest (Plunkett and Sundell 2013b; Belleri 2017, 2020; Thomasson 2017; Kouri Kissel 2021; Lewiński and Abreu 2022; Plunkett et al. 2023).

Much has been done, over the course of the past decade, to clarify the nature and manifestations of metalinguistic disagreements, as well as to motivate engineering and ameliorative projects issuing from such conflicts. As befits any philosophically interesting idea, however, new and different questions have opened up in the meantime. The contributions to this collection, individually described below, bring some of these questions into sharp relief; in so doing, they help advance extant discussions on four overlapping thematic fronts.

The first of these concerns the breadth (of applicability) of the notion of metalinguistic negotiation. Described in abstract and general terms, as we did above, the idea seems innocent enough. The flexible and capacious nature of the notion, moreover, as well as its promise of making good on the intuition that linguistic misalignments sometimes mask substantive disagreements, has been extensively advertised (Plunkett & Sundell). But everything – including theoretical conjectures and analyses – has a cost, and if this cost turns out to be very high, it may speak against the plausibility of an account. In the case at hand, a much-discussed worry is that the metalinguistic negotiation analysis of disputes may clash with independently motivated conceptualisations of linguistic agents’ interpretive competence (Abreu) as well as their semantic and pragmatic competence (Odrowąż-Sypniewska). Both criticisms may be seen as invitations to exercise methodological caution, reminding theorists to always keep an eye out for undesirable trade-offs.

Methodological caution, however, does not command inquisitive paralysis. Thus, the continuing investigation of the phenomenon (or phenomena) of metalinguistic disagreement remains well motivated, and in this sense it becomes increasingly pressing to clarify how exactly it is that metalinguistic disagreements can come about, how they come to be expressed and, eventually, be resolved. Notice that this question may be sharpened in more than one way. On one level of description, for instance, it can translate into the challenge of identifying the mechanisms of metalinguistic negotiation, at least in its more prominent and stable instantiations. A conversational implicature analysis seems to be a natural candidate (Odrowąż-Sypniewska), but it is perhaps not so clear that it will produce the appropriate predictions (Soria-Ruiz). On a different level of description, it can translate into the task of clarifying how different metasemantics create different possibilities for metalinguistic disagreement and conceptual engineering. Furthermore, and relatedly, it raises the question of how speakers’ underlying metasemantic views – their largely implicit views about what determines the meaning of their linguistic expressions – might influence which metalinguistic disagreements and disputes they engage in (De Brabanter & Leclercq).

Unearthing and paying close attention to linguistic agents’ metasemantic commitments, while unquestionably an important task, is also liable to reveal further deep-seated tensions between such commitments and the tenability of extant metalinguistic analyses of disputes. An especially prominent example of such tensions also motivates the title of this collection, as it motivated the eponymous conference we (guest editors together with Erich Rast) organised in Lisbon just over a year ago.

At its most general, semantic externalism has it that meaning is determined (partly or wholly) externally – ‘outside the head’, as it were (Putnam 1975; Burge 1979; Kripke 1980). The label is a broad one, capturing a wide gamut of views including more moderate and localised formulations of the core intuition. Even then, semantic externalism can plausibly be thought to raise special difficulties not only to the possibility of metalinguistic disagreements but also to the possibility of controlled and consequential metalinguistic intervention (see Cappelen 2018, Riggs 2019, Sawyer 2018, 2020, Koch 2021 and Pinder 2021 for discussion). If meaning is not determined by language users, then don’t metalinguistic disagreements ultimately boil down to someone being right and someone being wrong? And, similarly, doesn’t it follow that any attempt to engineer meaning change is ultimately futile?

More formally, it is not clear that verbal disputes can be as pervasive and stubbornly persistent as some authors claim, if the conditions are in place for stable and collectively shared meanings (Abreu). Nor is it clear that conceptual engineering can be effectively pursued if the metasemantic base is as inaccessible and unmanageably complex as some forms of semantic externalism suggest. At the very least, close attention must be paid to how external factors – partially or completely outside of our control – determine the applicability of our concepts (e.g. ‘woman’, ‘democracy’, ‘planet’, ‘Jewish’, ‘unbearable suffering’) and, consequently, the truth of some of our most personal and fundamental claims and judgements (Richard).

Finally, there is an intuitively important (if not fully transparent) sense in which metasemantics starts with speakers themselves. Accordingly, in addition to asking after the ways in which semantic externalism reflects on the possibility of metalinguistic disagreement, we can also enquire into how metalinguistic disagreements can help elucidate the semantics and metasemantics of terms, and into which brand of (meta)semantic externalism, if any, is true of them (De Brabanter & Leclercq).

Assuming that we remain duly vigilant when it comes to balancing metasemantic commitments and metalinguistic interpretations of disputes – and, perhaps, bracketing some or all of the previously mentioned worries about the latter – we may still reasonably wish to push ahead and see whether, and in what way, we can bring about meaningful change through metalinguistic intervention. Here, too, many details have yet to be filled in. Taking seriously the normative component of metalinguistic negotiations may certainly suffice as a motivation for conceptual engineering qua method. But taking seriously the idea of conceptual engineering, in turn, requires saying more: about how such interventions might actually work; when they should begin; how to negotiate stability and change – both at a principled, theoretical level (Belleri), and on the ground; and what they can tell us about, and how they might be shaped by, the part of the world they purport to target (Richard).

Against this backdrop, we now move to describe the individual contributions to the collection in more detail.

David Plunkett and Timothy Sundell’s “Reflections on Some Varieties of Metalinguistic Negotiation” offers a broad-scope perspective on the discussions that have unfolded over the past decade, driven in large part by their own contributions to the debate. Part of what the authors do in this paper, then, is review previous work on the idea of metalinguistic negotiation since its inception (Plunkett and Sundell 2013a). This is far from a perfunctory rehearsal of familiar material, however: even as they look back, Plunkett and Sundell’s ambition is to clear new paths ahead. Their purpose in this sense is at once clarificatory and constructive: on the one hand, to articulate key aspects of their original proposal in different, more explicit and illuminating terms; on the other, to expand on what they see as potentially fruitful new applications of the framework.

A central concern in their clarificatory effort is to highlight (what they take to be) the “capacious” nature of the phenomenon of metalinguistic negotiation. This is accomplished by identifying four broad, mutually interacting dimensions of variation of the target phenomenon, explored in as many sections. For instance, one level of variation tracks the possible motivations driving speakers to engage in a metalinguistic negotiation: whether these motivations are implicit or explicit, whether they are cooperative or rather adversarial in character, whether they are intended to serve goals that are (primarily) practical, theoretical, epistemic, prudential – each of these parameters will make a difference to the kind of metalinguistic negotiation that actually takes place.

Following this same line of thought, metalinguistic negotiations are shown to vary – actually or potentially – along three further dimensions: with respect to the intended audience(s) of the negotiation; the scope of the normative claims put forward in the dispute; the kind of normativity, and/or evaluative commitments, in play.

The resulting meta-level picture, Plunkett and Sundell maintain, vindicates the intuition that the phenomenon of metalinguistic negotiation is at once pervasive yet ultimately unified; and it brings into view the diverse capacities in which the notion itself, and its attendant theoretical framework, can be fruitfully employed in numerous areas and debates of philosophical interest – including, in particular, conceptual engineering.

A secondary aim of Plunkett and Sundell’s contribution is to resist worries, raised for instance by Shields (2021), as to the plausibility of their account of metalinguistic negotiation once one attempts to ‘fill in’ the details of recurrent illustrations of the phenomenon (e.g. the familiar “Waterboarding is (not) torture” scenario). Given that the authors’ aim is primarily constructive and forward-looking, it is understandable that they only engage with a small portion of the objections levelled at their account over the years. But one particular criticism – known as the speaker error objection – remains pressing for the metalinguistic analysis of disputes.

At its most general, the worry is that Plunkett and Sundell’s account seems to invite the problematic conclusion that linguistic agents engaged in metalinguistic negotiations are often confused about what they, and their fellow disputants, are doing. For, the account has it that disputants often operate under an assumption of adequate linguistic coordination – but mistakenly so, the idea continues, because the dispute actually takes place at a different level. Thus, while participants in the dispute interpret their conflict at face value – taking themselves to express the same (presumptively shared) meaning by their use of the same terms – the metalinguistic negotiation analysis has it that disputants mean different things by those same terms. And, while the former take themselves to be putting forward conflicting propositions about the world, the latter diagnoses the situation as a metalinguistic disagreement mediated by compatible object-level propositions.

The next two contributions in the collection offer additional glosses of this worry – which, according to both authors, has not yet been put to rest.

In “Metalinguistic Negotiation, Speaker Error, and Charity”, Pedro Abreu develops a novel version of the speaker error objection to the metalinguistic negotiation analysis (MNA) of disputes. Abreu’s contribution starts at the source of the problem itself: his first contention is that the MNA is committed to a form of speaker error – dubbed the Crucial Type of Speaker Error – that the literature has thus far failed to recognise, and that is especially worrisome for the account.

Having shown that the Crucial Type of Speaker Error is importantly distinct from variants previously examined by Plunkett & Sundell (2021b), Abreu moves to explain why it constitutes an especially problematic error ascription. The general argument is that, by interpreting against speakers’ own understanding of their disputes, the MNA is uncharitable in three different ways: first, it counts speakers as mistaken interpreters of their interlocutors; second, it counts speakers as uncharitable interpreters of their interlocutors; third, it counts speakers as mistaken interpreters of their own prior thoughts and utterances. Taken together, these three breaches of interpretative charity speak strongly against the MNA.

Abreu’s contribution ends with an interesting twist. The central argument in the paper is built on the assumption that, in many disputes recognizable as prototypical targets of MNA, speakers would reject that line of analysis. In the final section, conclusions are stated conditionally: if speakers show the right kind of resistance, then the MNA is problematic in the way described. And by the lights of arguments in the semantic externalist tradition (Burge 1979, Schroeter & Schroeter 2014), Abreu suggests, the antecedent does come out true in many relevant cases.

Joanna Odrowąż-Sypniewska also takes up a version of the speaker error objection in “Spicy, tall, and metalinguistic negotiations”, offering a different but congruous perspective to Abreu’s. Odrowąż‐Sypniewska seeks to press Plunkett and Sundell on two related points: the alleged pervasiveness of metalinguistic negotiations in all sorts of domains, involving all sorts of expressions; and the further tension between that claim and the independently plausible hypothesis that metalinguistic content is in many cases communicated pragmatically.

Plunkett and Sundell (2013a, 2021b) acknowledge that speakers may often (be disposed to) reject the theoretical characterisation of their disputes as metalinguistic, and try to show that this does not constitute an especially worrying difficulty for their view. Odrowąż-Sypniewska sees it differently, arguing that the kind of speaker error to which Plunkett and Sundell’s account is committed – construing speakers as unable to reliably distinguish semantic from pragmatic content – is far more serious than they acknowledge. The problem is not a confusion between different levels of content, but rather speakers’ tout court rejection of any metalinguistic content, be it at a semantic or at a pragmatic level.

Finally, Odrowąż-Sypniewska drives her point home by showing that the sort of incompetence implicitly attributed to linguistic agents is irreconcilable with a Gricean explanation of how metalinguistic negotiations might actually come about.

Notice that there is a double-layered challenge to Plunkett and Sundell’s account here. On the one hand, both these authors have repeatedly emphasised their neutrality with respect to the question of which mechanism(s) might be thought to underlie the metalinguistic negotiation phenomenon. Such neutrality, however, is presumably unsustainable in the long run: if metalinguistic negotiations occur at all, and a fortiori if they occur as frequently as they are alleged to, they must occur somehow. On the other hand, the hypothesis that pragmatic mechanisms – in particular, implicature mechanisms – are involved in the negotiation of metalinguistic content seems prima facie plausible. But if linguistic agents are indeed as systematically prone to misidentifying the content of their utterances, it’s far from clear that they could convey such content via implicature. What is more, it’s even less clear that this sort of radical incompetence could be reconciled with the independently plausible idea that linguistic agents convey all sorts of content via implicature, in a large portion of their communicative lives.

One might of course draw an entirely different moral from the foregoing: if speaker error objections derive at least part of their force from the supposition that metalinguistic negotiation involves exchange of implicated content, then a natural thought is to look for an alternative explanatory analysis of the phenomenon.

One such alternative is presented by Andrés Soria-Ruiz, whose contribution “Is metalinguistic usage conversational implicature?” mounts a contrastive defence of a neo-Stalnakerian account of metalinguistic usage against implicature analyses.

The paper begins with a comprehensive critique of what is to date the most detailed formulation of an implicature account of metalinguistic negotiation, put forward by Mankowitz (2021). Soria-Ruiz observes, first, that metalinguistic propositions and conversational implicatures respond differently to characteristic cancellability, reinforceability, and embeddability tests. He then argues that Mankowitz’s account both under-generates and over-generates metalinguistic interpretations; in so doing, Soria-Ruiz notes that the implicitly favoured reading of the relevant implicatures in substitutional terms may constitute a further difficulty for the account. Combined, these objections put pressure on proponents of pragmatic analyses to improve the predictive capacity of their accounts of metalinguistic usage.

In the more positive part of the paper, Soria-Ruiz takes his cue from work by Barker (2002, 2013) to outline a broadly Stalnakerian (1978) account of metalinguistic usage. This is presented as a semantics based on world-interpretation pairs, on which sentences receive both an object-level and a metalinguistic reading depending on the context of utterance. The account’s in-built context-sensitivity, Soria-Ruiz argues, allows it to better capture certain key properties of metalinguistic usage compared to extant implicature analyses.

Next, in “From Semantic Deference to Semantic Externalism to Metasemantic Disagreement” Philippe De Brabanter and Bruno Leclercq focus their attention on a specific kind of metalinguistic disagreements: metasemantic disagreements, understood as disagreements that come about as a result of underlying conflicts between speakers’ metasemantic views – views about what ultimately determines the meaning of linguistic expressions. The authors produce a fine-grained typology of their target phenomenon, showing that it exhibits a perhaps unexpected level of complexity and variation.

The idea that ordinary speakers may entertain metasemantic views plays a pivotal role in the paper. Especially interesting is the authors’ proposal that these same views may be tracked by, and extrapolated from, speakers’ deferential attitudes with respect to particular expressions. Among the metasemantic views identified in this way, variations on the externalist theme take centre stage. Notably, De Brabanter and Leclercq’s emphasis on speakers’ ability to choose between different types of deferential attitude suggests a certain freedom to determine which facts ground our meanings. This is reminiscent of Ludlow’s (2014) argument for the compatibility between (meta)semantic externalism and “meaning control”. It is also a controversial position – perhaps most famously rejected by Cappelen (2018).

De Brabanter & Leclercq’s contribution thus presses for serious engagement with the question: How should ordinary speakers’ views as to the facts that ground meaning shape theoretical inquiry into metasemantics and into conceptual engineering? This question underpins a broader, ongoing discussion in the literature (see also Abreu, Odrowąż-Sypniewska), to which the present paper adds important new elements.

Conceptual engineering takes centre stage in the remaining two contributions to the collection. In “Conceptual Engineering Between Representational Skepticism and Complacency: Is There a Third Way?”, Delia Belleri addresses the problem of when to start conceptual engineering: What should our attitude towards our concepts be? What should prompt us — qua regular linguistic agents or qua conceptual engineers — to even start evaluating, and possibly revising, our concepts?

Belleri has an original answer to this question: critical concept conservatism, a complex attitude towards received concepts that combines a default element of concept conservatism and a critical element that is activated whenever reasons for doubt and potential change are recognized as stronger than reasons for stability.

Belleri’s rich discussion presents a variety of considerations in support of critical concept conservatism, both positive and negative. On the one hand, she argues that it constitutes an attractive alternative to two untenable positions, labelled representational complacency and scepticism and recently outlined by Cappelen (2020). On the other, she maintains that critical concept conservatism enjoys desirable affinities with belief conservatism. Of particular note is Belleri’s argument that entrenchment — the concept’s already being there, and having been there for a fairly long time — combined with a reputable career, can constitute pro tanto reasons for retaining a concept. Crucially, however, the possibility of critical assessment and revision is never out of reach for the critical concept conservative.

Critical concept conservatism is described, and defended, as a general, “broad-spectrum” view. Naturally, then, the specific details that will tip the scales towards revision or retention in any given case are not pre-packaged into the view; but the story can easily be adapted to account for particular cases, Belleri maintains. The point of the paper is to equip us — concept users and engineers — not so much with a ready-to-use tool, as with a principled guide for striking a reasonable balance between two opposing drives.

Knowing when to start conceptual engineering, while important, is likely not enough to guard aspiring innovators against trouble further downstream. Any attempt to ameliorate a concept will fall flat if it is not supported by at least some segment of the linguistic community and – more broadly – the social context(s) we inhabit. And, as Mark Richard warns in “Conceptual engineering: Be careful what you wish for”, the absence of a reliably supportive socio-normative context may compromise our ability to speak truly, when using (or attempting to use) an improved concept that deviates (there and then) from the dominant semantic paradigm.

Such is the case, and such are the dangers, with the concept woman, as with the concept Jewish, and with concepts picking out socially constructed kinds more generally, Richard maintains. How so? Consider, as Richard does, gender identity concepts. There is a straightforward case to be made for taking gender kinds to be socially constructed. But then it is all but straightforward – at least prima facie – to see what we should make of, say, a dispute between a trans woman, who says of herself that she is a woman, and an anti-trans activist, who denies that this is so. More fully: it is not immediately clear that we can say, about a dispute of this sort, that the trans woman’s claim is both normatively correct (in the present-day situation, where successful inclusionary meaning change is still a distant goal) and descriptively correct (even prior to successful engineering). For, prima facie, we seem to be pulled either towards relativism – to conclude that, say, the trans woman’s claim is true for her but false for her interlocutor – or towards stasis. Conceptual engineers thus face a challenge; to escape it, Richard argues, they would do well to take on board a Putnam-style brand of pragmatic realism.